Twenty-three years ago, an SEO specialist packed keywords into progressively longer phrases, like Russian nesting dolls. "Airfare to Philadelphia" contained "cheap airfare to Philadelphia," which contained "cheap airfare to Philadelphia today." A single piece of content was meant to catch multiple searches, from short-tail to hyper-specific.
In 2026, the data tells us something few would have predicted: artificial intelligence generates sub-queries in nearly identical patterns. A recent article on Search Engine Journal revisits this two-decade-old strategy. And from our perspective at difrnt., a Romanian digital marketing agency, it deserves serious attention precisely because it isn't new.
Why does a 23-year-old tactic matter now? Because the industry spent the last decade optimizing in the wrong direction. We focused on short head terms with high volume, while users (and now AI) searched with increasing specificity. The recent data from both independent researchers and Google itself confirms exactly this trend.
Russian dolls in SEO: short phrases inside long ones
The concept is straightforward. Instead of optimizing a page for just "flight tickets," you build content around the 4-5 word variant: "cheap flight tickets Bucharest to London." The page automatically ranks for the shorter version, but you also capture the more specific long-tail queries.
Why does this work? Because 15% of daily Google searches are completely new. Google's John Mueller confirmed this figure in March 2025 at Search Central Live in New York, noting it has remained surprisingly consistent over the years.
These new queries come from daily context, from word combinations nobody has used in that exact form before. Nested phrases make the difference here: your content covers combinations you never anticipated, because the structure of longer phrases naturally contains the shorter ones. We explored something similar when we wrote about the keyword universe that was always smaller than we thought.
AI thinks in long phrases. The data confirms it.
The most compelling data comes from a study published in August 2026 by MJ Cachón. She ran 189 branded prompts through ChatGPT and found the system generated 1,797 sub-queries, averaging 7 words each. Users never typed any of them. The AI created them on its own, breaking down the initial question into progressively more specific variants.
Even more telling: quote usage climbed 25-fold between the first sub-query and the last. In other words, AI starts with a general question and progressively narrows toward exact phrases, searching for textual confirmations from source material.
Google has confirmed the trend from their side too: AI Mode queries run three times the length of traditional search queries. We're no longer talking about 2-3 word keywords. We're talking about full, contextual, specific phrases. If your content contains only generic terms, it simply won't appear in this type of search. The gap between how people actually ask questions and how most websites describe their services is growing wider with every AI update.
Three habits that make a practical difference
Look for nested phrases, not just the head term. Open Google Search Console and check queries with high impressions but low clicks. That's where you'll often find long variants of your main keywords: 4-7 word phrases your page partially covers. Build content around these longer variants. We've explored what Search Console sees and doesn't see from AI traffic, and this approach perfectly complements that analysis.
Publish at news speed, not editorial calendar speed. Press releases, rapid-response articles, trend commentary: they all have an advantage that a blog post scheduled for next month lacks. They're the first to capture language around a new event. When an algorithm change or a new tool drops, whoever publishes first captures the phrases others haven't even identified yet. This doesn't mean sacrificing quality for speed. It means having an internal process that enables rapid publication when a real opportunity appears.
Write sentences that can be quoted on their own. AI verifies information by searching for exact passages in sources. If your answer to a question is buried in a long, ambiguous paragraph, AI won't cite it. But if you write a clear sentence that directly answers a question ("Product feeds need to be updated at least every 24 hours to appear in ChatGPT Shopping"), that sentence becomes quotable.
What this means for your website
You don't need to be an SEO specialist to apply what we've described above. Here's the essential point: content on your site needs to answer specific questions, not just describe services generically. A "SEO services" page that says "we offer comprehensive optimization services" will never appear in an AI response. A page that explains "how a technical SEO audit works for an e-commerce site with over 10,000 products" has considerably better chances.
Speed matters too. Not site speed (although that's important as well), but how quickly you react to industry changes. Search behavior is changing, and every day you delay publishing is a day someone else captures the new phrases.
The 2003 technique isn't a nostalgic trick. It's a principle for structuring content that has proven valid for both classic search engines and AI systems. The difference is that now, with AI Mode and tools like ChatGPT Shopping, long phrases aren't optional. They're the primary way AI finds and verifies information. And that might be the most important lesson here: the best strategies aren't always the newest. Sometimes they're the ones that withstand the test of time and adapt naturally to a new context.
